DEPARTMENT OF ECONOMICS WORKING PAPER SERIES
Declining Effects of Oil Prices
Munechika Katayama
Louisiana State University
Working Paper 2009-02
http://www.bus.lsu.edu/economics/papers/pap09_02.pdf
Department of Economics
Louisiana State University
Baton Rouge, LA 70803-6306
http://www.bus.lsu.edu/economics/
Declining Effects of Oil-price Shocks
Munechika Katayama
∗Louisiana State University
mkatayama@lsu.edu
This Version: January 19, 2009
First Draft: May, 2007
Abstract
Output responses to oil-price shocks not only tend to be weaker, but also to peak
earlier recently. This paper builds a model that incorporates a realistic structure of
US petroleum consumption and explores three possible explanations for the changes.
The first is based on deregulation in the transportation sector, which has brought more
competition and improved efficiency in the industry. The second is overall
improve-ments in use of energy. The third is less persistence of the oil-price shock. Under
realistic parameter values, it is demonstrated that all three factors play an important
role quantitatively. These three factors together could account for a 51% reduction in
the peak response of output to an oil-price shock.
Keywords: Oil-price shocks; recessions; deregulation.
JEL Classification: E20, E32, L43, Q43.
∗I am grateful to James Hamilton for his encouragement and helpful comments. I have also benefited
from suggestions from Takeo Hoshi and Valerie Ramey, and from discussions with Kwang Hwan Kim and Yong-Gook Jung. I also thank to seminar participants at UCSD and Louisiana State University.
1
Introduction
Macroeconomic consequences of large increases in the price of oil have been of great concern
among economists and policy makers, as well as the general public since two major oil-price
shocks hit the economy in the 1970’s. In recent years, however, it seems that the effect of
oil-price shocks has been decreasing. For example, the Congressional Budget Office reports
that:
Contrary to general expectations, the large and persistent rise in energy prices
that has occurred over the past two and a half years has not caused substantial
problems for the overall U.S. economy. Although many households have had
trouble adjusting to the higher prices, the effects on the nation’s gross domestic
product (GDP), employment, and inflation have thus far been moderate.
(Con-gressional Budget Office, 2006, p.VII)
It is well known that, after World War II, nine out of ten recessions in the US economy
were preceded by large increases in the oil price. A general perception is that oil-price
increases may cause a prolonged and deep recession. In fact, a number of studies have
tested the relation between economic activity and oil prices and have confirmed that it is
not a statistical coincidence.1 There have been substantial hikes in the oil price in the early
2000’s. A couple of events, such as the civil unrest in Venezuela in 2002, the Iraq war of
2003, and Hurricanes Katrina and Rita in 2005, triggered increases in the oil price. During
2006, the nominal oil price has reached value more than three times as high as the average
price in the 1990’s. However, these large oil-price hikes did not result ins significant economic
downturns.2
1Those studies include Hamilton (1983), Burbidge and Harrison (1984), Gisser and Goodwin (1986),
Daniel (1997), Carruth, Hooker, and Oswald (1998), and Hamilton (2003). However, Bernanke, Gertler, and Watson (1997) argue that oil-price shocks were not a major cause of economic downturns. See also subsequent discussions in Hamilton and Herrera (2004) and Bernanke, Gertler, and Watson (2004). Barsky and Kilian (2002, 2004) also raise a skeptical view about the role of the oil-price shocks in recessions in the 1970’s.
In order to see the possible changes in the US economy over time, Figure 1 plots
differ-ent behavior of macroeconomics variables after a 10% increase in the net oil price increase
(NOPI) of Hamilton (1996, 2003). These impulse response functions describe average
behav-ior following a 10% increase in oil prices above their previous 3-year high. Sample periods are
split into two sub-samples at 1984, following McConnell and Perez-Quiros (2000). As can be
seen clearly, we can observe different patterns in response to the oil-price shock, measured
by an increase in the NOPI, in two sub-samples.3 Not only output but also other variables
display weaker responses to the oil-price shock in the post-1984 period. This is very robust
to different specifications. Furthermore, the peak of the output responses appears earlier in
the post-1984 sub-sample than the pre-1984 sub-sample. On average, the output responses
peak at 6 quarters after the shock with a 1.52% decline compared to the unshocked path in
the pre-1984 sub-sample, whereas, in the post-1984 periods, the responses peak at 3
quar-ters after the shock only with a 0.38% drop from the unshocked path. The peak response of
output declines by 75% and the peak is shifted by 3 quarters. This suggests that there have
been some changes in the economy.
This paper takes the evidence presented in Figure 1 seriously and explores possible
explanations why we expect to observe the weaker responses of macroeconomic variables
to the same size of the oil-price shock. Most of existing studies have focused mainly on
recessionary consequences of oil-price shocks (e.g., Kim and Loungani, 1991; Rotemberg and
Woodford, 1996; Finn, 2000; Leduc and Sill, 2004; Carlstrom and Fuerst, 2006;
Aguiar-Conraria and Wen, 2007; Cavallo and Wu, 2007) and have not addressed the issue of weaker
effects of oil-price shocks. The only exception is Blanchard and Gal´ı (2007), who ask the same
question as this paper does. In order to answer the question, they investigate four different
hypotheses; (a) good luck (i.e., lack of concurrent adverse shocks), (b) smaller share of oil
in production, (c) more flexible labor markets, and (d) improvements in monetary policy.
They conclude that all four factors have played an important role in accounting for the mild
3Early studies, such as Mork (1989) and Hooker (1996), also document that, as the sample period is
GDP 5 10 15 20 −2 −1 0 1 2 Consumption 5 10 15 20 −2 −1 0 1 Inflation 5 10 15 20 −1.5 −1 −0.5 0 0.5 1 1.5 2 Investment 5 10 15 20 −10 −5 0 5 Real Wage 5 10 15 20 −1 −0.5 0 0.5 Labor Productivity 5 10 15 20 −1.5 −1 −0.5 0 0.5 1 Interest Rate 5 10 15 20 −100 −50 0 50 100 Pre−1981 Post−1984
Figure 1: Different Responses to the Oil-price Shock: Before and After 1984
Note: Each panel plots responses to a 10% increase in NOPI. These impulse responses are obtained from 8-variable VAR with 4 lags, which is similar to the specification in Christiano, Eichenbaum, and Evans (2005). Variables included are NOPI, real GDP, real consumption, GDP deflator, real investment, real wage, labor productivity, and interest rate. A constant term is included as well. The sample starts from 1951:Q1 and ends at 2007:Q4. Except for NOPI, all variables are logged and retrieved from the FRED database of the FRB St. Louis. Horizontal axes indicate quarters and the vertical axes take percentage deviations from the unshocked path. Shaded areas represent 95% confidence bands for the pre-1984 IRFs, while dashed lines show those for the post-1984 IRFs.
effects on inflation and economic activity.
Besides the hypotheses Blanchard and Gal´ı (2007) have considered, there are also other
possibilities for the reason why we might expect the effect of the oil-price shock on economic
activity has become weaker. In this paper, I will analyze three possible explanations for the
declining effect of oil-price shocks in detail. The first factor investigated is the outcome of
macroeco-1950 1960 1970 1980 1990 2000 0 1000 2000 3000 4000 5000 6000 M il li ons Ba ll el s 1950 1960 1970 1980 1990 2000 2000 3000 4000 5000 6000 7000 8000 Petroleum Consumption by Sector (Millions Ballels)
1950 1960 1970 1980 1990 2000 10 20 30 40 50 60
Petroleum Consumption Share by Sector (%)
Total (Right Scale)
Industrial Residential Electric Power Transportation Industrial Electric Power Residential Commercial Transportation Commercial
Figure 2: Petroleum Consumption by Sector 1949–2005
Sources: Department of Energy, Energy Information Administration (2006), Annual Energy Review 2005, Table 5.13a–5.13d.
nomic consequences of an oil-price shock, we usually have the industrial/commercial sector
in mind in terms of petroleum consumption. It is always helpful to look at data first to
see how petroleum is used across sectors. Figure 2 shows petroleum consumption in the US
economy from 1949 to 2005 obtained from the Energy Information Administration (EIA). It
reveals that contrary to common treatment in the existing models, more than a half of the
US petroleum consumption is coming from the transportation sector. Thus, it suggests that
it is important to pay more attention to the role of petroleum consumption in
transporta-tion. Commercial transportation had been one of the most regulated industries in the US
economy. This regulatory environment had created inefficiencies and economic rents, and
deregulation resulted in lower costs and a more competitive environment. In this paper, in
addition to the typical industrial sector, I explicitly model the role of the transportation
industry in order to account for the effect of the deregulation in the transportation industry
that began in 1980.
Another possibility I will examine is the effect of improvements in energy efficiency and
oil-price shocks in the 1970’s, technological advances enabled us to utilize petroleum more
efficiently. For example, vehicles’ miles per gallon improved dramatically. According to the
Energy Information Administration (2006), from 1973 to 1991, miles per gallon (all motor
vehicles) improved by 42%. This more efficient use of oil is apparent in terms of the oil
expenditure share in value added. While average oil expenditure share relative to GDP was
3.65% in the pre-1984 periods, the average share declines in the post-1984 periods to 2.75%.
This point is similar to one of factors considered in Blanchard and Gal´ı (2007). Since there
exist two types of oil usage in the model, industrial sector and transportation sector, it is
possible to analyze the consequence of improved energy efficiency in more detail. In this
sense, this paper is complimentary to Blanchard and Gal´ı (2007).
Lastly, I also consider a possible effect of changes in persistence of the oil-price shock.
Small changes in the underlying shocks process could alter the responses of macroeconomic
variables. Figure 3 plots real price of oil measures by the PPI of crude petroleum deflated
by GDP deflator. Judging from data, it is clear that the oil-price behavior shows different
patterns after the mid 1980’s. Particularly, it tends to be less persistent compared to the
near-unit-root nature of the oil price we have observed up until the mid 1980’s. The goal of
the paper is to investigate the role of these three factors in accounting for the two changes
observed in the time series data, that is, weaker responses of output and the earlier timing
of the peak.
The model is based on a standard model with adjustment costs. I extend the
stan-dard sticky-price model to include the role of oil prices into the economy in two ways. As
in Finn (2000), variable capital utilization tied to petroleum consumption is introduced in
the intermediate-good production process. Another extension is to include a transportation
sector as an additional sector in the economy. It is assumed that all produced
intermedi-ate products must be delivered to customers by using transportation services. Including
transportation firms amplifies the effects of the oil-price shock, because, as will be described
1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Figure 3: Real Price of Oil
should be noted that, although there are empirical studies documenting possible asymmetric
effects on the economic activity (see, for example, Davis and Haltiwanger, 2001; Hamilton,
2003)4, like other existing models, the structure of the model considered here implies that
increases and decreases in the price of oil would have symmetric effects. However, the focus
of this paper is on the recessionary consequences following a large increase in the price of oil
and on accounting for the changes in the economy over time.5
The baseline specification of the model, which corresponds to the pre-1984 state of the
economy, well captures important aspects of the economy’s response to the oil-price shock,
under reasonable parameter values. In terms of accounting for the changes in the economy
overtime, each factor individually could account for about a 20-25% decline in the peak
response of aggregate value added. Combined together, the three factors result in a 51%
reduction in the peak response and shift the timing of the peak by 2 quarters, compared
4Although the asymmetric effects of oil-price shocks are widely believed, Kilian (2007) summarizes
sup-porting evidence for the symmetric responses of monthly real consumption expenditure based on formal statistical tests and suggests that nonresidential investment could be a source of the asymmetry.
5However, taking the transportation industry into account may help resolve the inability of modeling
asymmetric effects since a common practice in the industry is to include fuel surcharges in invoices, when the economy encounters rapid increases in fuel costs. Fuel surcharge, by construction, accrues when the oil price increases, but it is not applied when the oil price decreases. That is, potentially the structure of the industry itself is asymmetric in terms of responses to oil-price shocks.
with the baseline specification.
The rest of the paper is organized as follows. Section 2 provides some background
infor-mation on regulation in the US transportation industry, as well as petroleum consumption
of the transportation sector. Section 3 presents a model that we can use for assessing the
three possible factors. Results are in Section 4 and Section 5 concludes.
2
Transportation Industry Background
2.1
Petroleum Consumption for Transportation
As mentioned in the Introduction, most of US petroleum consumption comes from the
trans-portation sector. One caveat is that the EIA’s definition of the transtrans-portation sector includes
the supply-side petroleum consumption as well as petroleum use of households for driving
automobiles. The EIA does not have data on how much petroleum can be attributed to
fuel of household vehicles except for selected years. Based on the information in the EIA’s
“Household Vehicles Energy Use: Latest Data & Trends” 6, about 52% of petroleum is used
by household vehicles in the transportation sector after the late 1980’s. Therefore, even when
we take account of household vehicle use, the transportation service industry itself consumes
as much oil as the whole industrial sector does.
Figure 4 displays the share of oil consumption in the transportation sector by mode.
As can be seen clearly, within the transportation industry, the trucking industry is a major
source of oil consumption. Any investigation of the recessionary consequences of oil-price
shocks needs to pay particular attention to the trucking industry.
2.2
Regulation in the Transportation Industry
The U.S. transportation industry in the 1970’s was heavily regulated by the Interstate
Truck Air Marine Rail 0 10 20 30 40 50 60 70 80
Percent of Freight Petroleum Consumption
1985 1988 1991
Figure 4: Petroleum Consumption in the Transportation Industry by Mode
Source: Department of Energy, Energy Information Administration (1995), Measuring Energy Efficiency In The United States’ Economy: A Beginning, Figure 5.17
The ICC regulated prices and competition in interstate transportation. Starting from
rail-roads, the control of the ICC was expanded to a variety of transportation modes, such as
trucking, water transportation, oil pipeline, and freight forwarders. The trucking industry is
likely to be a good example of a perfectly competitive industry, if there is no regulation in
the industry (Olson, 1972). Such a regulatory environment in the transportation industry
imposed by the ICC had reduced competition and created significant inefficiency. However,
the regulatory environment came to an end in 1980. The Motor Carriers Act of 1980 and the
Staggers Rail Act of 1980 triggered deregulation in the transportation industry. Throughout
the 1980’s, the control of the ICC had been diminishing. Ultimately, all of regulations were
removed in 1995 when the ICC was abolished.
Regulations by the ICC primarily consisted of restrictions on operating rights and a
requirement of filing price changes. The ICC controlled new entries by granting operating
certificates, which specified a scope of operation, such as a list of routes, commodities,
the new service contributed to the public convenience and necessity and that incumbent
carriers were unable to provide the service. New applications were frequently protested by
incumbent carriers. This procedure was also applicable to incumbents expanding their scope
of operations, such as more variety of commodities and operation routes. Such a difficulty
led many firms to purchase the operating authorities of other carriers. The market value of
those operating rights could be a measure of the monopolistic rents. In fact, those operating
rights were listed as assets on regulated firms’ books. Snow and Sobotka (1977) report
that the market value of the operating rights amounted to 15-20% of gross annual revenues.
The ICC’s control of operating rights created not only market power, but also inefficiencies.
Restrictions on routes and products often resulted in empty backhauls. For example, based
on the 1976 ICC study, Wright (1983) argues that about 38% of interstate and 71% of
intrastate backhauls were empty.
In addition to the entry control, regulated carriers were required to file all changes in
their rates with the ICC, thirty days prior to the new rate becoming effective. Within thirty
days, the proposal of a new price was subject to objections from anyone, including the ICC
itself and competitors (both those in the same transportation mode and from different types
of carriers). Once the proposal was protested, the ICC would investigate its reasonableness.
Ratemaking and filing to the ICC were usually done in a collective manner, via regionally
distributed rate bureaus. Those rate bureaus approved by the ICC had been operated
under antitrust immunity since 1948 (the Reed-Bulwinkle Act). Although the ICC required
rate bureaus to be open and democratic, investigations done by the U.S. congress and the
ICC revealed that “these rate bureaus are, for most part, dominated by small groups of
large carriers” (U.S. Congress, Senate, 1980, p. 60). Due to limited resources of the ICC
reviewing proposed tariffs,7 virtually it was the case that “[t]he carriers set their own rate”
(U.S. Congress, Senate, 1980, p. 80).
7For example, according to the report from U.S. Congress, Senate (1980), 278,039 tariff publications were
filed to the ICC in 1978. On average, 5,000 pages of tariffs are filed daily. However, there were only 60 employees at the ICC who were in charge of checking those tariffs.
The rate bureau system ensured that protestants could impose substantial costs on
applicants trying to underprice other carriers, and hence created an incentive not to lower
prices. An individual carrier could propose selective rate changes. Such proposals were
usually forwarded to the standing rate committee, which set up a public hearing on the
proposal. Any interested parties could participate, regardless of whether they were directly
affected by the proposal. Frequently, those proposals were opposed by other carriers,
some-times causing the ICC to investigate the proposed rate, and making rate changes extremely
time-consuming and difficult. The ICC might then suspend the rate and call for a formal
investigation of its legality. It was possible for an individual carrier to file an independent
action to the ICC without going through the review at the rate bureau. However, most of
rate bureaus required the independent action to be informed before taking effect and other
carriers had an option of whether to join in the new rate or not. Thus, the rate bureau system
under the ICC regime created an incentive not to lower prices and discouraged competitive
pricing in the industry.
These regulations and controls by the ICC had created oligopolistic environment in the
transportation industry. For example, Winston (1981) estimates the markup over marginal
cost for various commodity categories. Among regulated commodities, the estimated markup
rates range from 29% to 64% (with the mean of 52.2%) for railroads and those for trucking
range from 26% to 49% (with the mean of 37.1%).
The deregulation resulted in freer entries and more flexible expansions of the scope of
business. The Motor Carriers Act of 1980 eliminated entry barriers. It required the ICC to
grant operating rights to any requesting firm that is able to provide the service. The burden
of proof was reversed. Although, prior to the Motor Carriers Act of 1980, new entrants had
been required to prove the necessity of new service, opponents were asked to provide evidence
why the new service was not beneficial. Furthermore, route restrictions were eliminated as
well. By 1986, the number of motor carriers holding ICC operating certificates was more
The Motor Carriers Act of 1980 also increased price flexibility. It guaranteed individual
motor carriers and freight forwarders complete freedom to change rates as long as price
changes are within the ±10% range compared with the level one year prior to a specified point
of reference. The ICC could increase the range up to 15% for rate increases in any year that
it finds that there is sufficient competition to regulate rates, and that the carriers, shippers
and the public would derive a benefit from the increased flexibility. The law prohibited the
use of the rate bureau for single-line rate making. Although the rate bureaus themselves
were not abolished, expanded independent actions increased pricing flexibility and reduced
importance of the rate bureau system.
Overall, the net effect of the deregulation can be summarized as higher efficiency and
more competition, which led to lower average costs. Ying and Keeler (1991) estimate the
ef-fect of deregulation on rates charged by motor carriers, controlling for various factors, such as
changes in primary factor prices and fuel costs. They find that, on average, prices decreased
by 15-20% within three years after the deregulation and that prices further dropped by
25-35% by 1985. Nebesky, McMullen, and Lee (1995) estimate and compare the markup rate
of the less-than-truckload segment of the trucking industry before and after the
deregula-tion. They find that the estimated markup has declined dramatically after the dereguladeregula-tion.
While the estimated markup rate using a 1977 sample is 35.5%, the estimate based on 1988
data is 1.9%.
In terms of the changes in the economy over time, I will interpret the consequences
of the deregulation as (i) lower markup due to the more competitive environment and (ii)
improved energy efficiency in the transportation industry. The following section will present
3
The Model
The economy consists of continuums of households, intermediate-good firms, and
transporta-tion firms, which are indexed by h ∈ [0, 1], i ∈ [0, 1], and n ∈ [0, 1], respectively, as well as
the monetary authority and the aggregator of final-good and labor services.
Each household supplies different types of labor services, which are imperfect substitutes,
in a monopolistically competitive market. There exist nominal wage rigidities in the form of
quadratic wage adjustment costs.
While intermediate-good firms produce distinct products and sell their output in
mo-nopolistically competitive markets, they act as a price-taker in factor markets. Production
of intermediate goods requires two additional inputs, other than capital and labor. First,
I assume that all produced goods must be delivered to consumers and intermediate-good
producers incur the cost of transportation services. Second, the firms also use raw materials
as one of factor inputs. The raw materials are assumed to be composed of all intermediate
goods. That is, produced intermediate goods will be used to assemble the final good for
households’ consumption as well as used as materials inputs for the production process.
Transportation firms supply distinct transportation services to intermediate-good firms
in monopolistically competitive markets. Their services are distinct in the sense that there
exist segmentations in the US transportation industry, such as operating region, products
carried, and transportation modes.
There are two direct avenues whereby an increase in the price of oil could have impact
on the economic activity. One is variable capital utilization tied with use of energy, as in
Finn (2000), Leduc and Sill (2004), and Cavallo and Wu (2007). This is intended here to
capture the role of the industrial sector in US petroleum consumption mentioned earlier.
The other channel through which the oil-price shock may have an effect in the economy is
direct impact on the cost of transportation. In the subsections below, we shall see the details
3.1
Aggregators
There are two types of aggregators in the economy. The final-good aggregator purchases
differentiated goods from intermediate-good firms and assembles the final good, which is
sold to households for consumption in a competitive market. The labor-service aggregator
hires differentiated labor services from households and transforms them into the composite
labor input, which is demanded by both intermediate-good firms and transportation firms
in a competitive market.
Both aggregation technologies are based on a typical constant returns to scale CES
technology. The final good and the composite labor are produced by
Yt= Z 1 0 Yt(i)(θ−1)/θdi θ/(θ−1) , Nt= Z 1 0 Nt(h)(θN−1)/θNdh θN/(θN−1) ,
respectively, where Yt(i) represents the gross output of intermediate-good firm i, Nt(h) is the
labor service of household h, and θ > 1 and θN > 1 are the elasticity of substitution between
the different inputs.
Taking the price of inputs as given, the aggregators minimize total costs. The first-order
conditions give us constant-elasticity inverse demand functions for intermediate products and
labor services Yt(i) = Pt(i) Pt −θ Yt, Nt(h) = wt(h) wt −θN Nt,
respectively, where Pt(i) is the price of intermediate good i, wt(h) is real wage for labor
ser-vices supplied by household h, and Pt =
h R1 0 Pt(i) 1−θdii1/(1−θ)and w t = h R1 0 wt(h) 1−θNdh i1/(1−θN)
3.2
Households
There is a continuum of households indexed by h ∈ [0, 1]. Households are identical except
for the heterogeneity of labor. Each household maximizes the expected utility
E0 ∞ X t=0 βt log (Ct(h)) + (Mt(h)/Pt)1−σ 1 − σ − Nt(h)1+ϕ 1 + ϕ ,
where 0 < β < 1 is the subjective discount factor, Ct(h) is consumption of final goods,
Mt(h)/Pt is real money balances, and Nt(h) is hours worked.
Each household carries Mt−1(h) units of money and Bt−1(h) bonds into period t. In
addition to the carry-over, the household receives a lump-sum transfer Tt(h) from the
mon-etary authority at the beginning of each period. After the maturity of bonds, the household
also receives Bt−1(h) additional units of money. During the period t, the household provides
labor in exchange for labor income. The household also receives nominal dividends from
intermediate-good producers and transportation firms, denoted by Πt and ΠTt, respectively.
In addition, each household is a monopolistic supplier of differentiated labor services
Nt(h) and faces a downward-sloping labor demand curve. Each household sets its own
nominal wage subject to quadratic wage adjustment costs, as in Kim (2000). The wage
adjustment cost function takes the form
ACW t (h) = φW 2 Wt(h) Wt−1(h) − πs 2 Wt(h) Pt ,
where φW > 0 is a wage adjustment cost parameter, Wt(h) is household h’s nominal wage,
πs is the steady-state rate of inflation, and Pt is the general price index. Thus, household h
faces the following budget constraint:
Mt−1(h) + Tt(h) + Bt−1(h) + Wt(h)Nt(h) + Πt(h) + ΠTt(h) Pt = Ct(h) + Bt(h)/rt+ Mt(h) Pt + ACW t (h), (1)
where rt is the gross nominal interest rate.
3.3
Intermediate-good Firms
There is a continuum of intermediate-good producers indexed by i ∈ [0, 1]. Each intermediate
firm produces a distinct product eYt(i) and sells it in a monopolistically competitive market.
Unlike typical models, I assume that all goods produced must be delivered to customers and
each firm incurs transportation costs. Specifically, each intermediate-good producer must
purchase transportation services Tt(i) in order to finalize its production process. We assume
that the whole production process is described by
e Yt(i) = min Qt(i) 1 − χY ,Tt(i) χY , (2)
where χY ∈ (0, 1) is a technology parameter, which determines complementarity of
trans-portation services, Qt(i) is firm i’s gross output, and Tt(i) is firm i’s demand for
transporta-tion services, which will be defined below.
Following Rotemberg and Woodford (1995) and Basu (1996), we assume that
produc-tion of gross output involves use of material inputs through the fixed-coefficient technology
between value added Vt(i) and material inputs Xt(i), given by
Qt(i) = min Vt(i) 1 − χQ ,Xt(i) χQ , (3)
where χQ ∈ (0, 1) is a share of materials cost in the value of gross output. Each firm produces
its value added Vt(i) by using capital services ut(i)Kt(i) and labor input NtI(i) with a typical
Cobb-Douglas production technology
Vt(i) = [ut(i)Kt(i)]α[NtI(i)] 1−α
, (4)
The material inputs Xt(i) are the CES aggregate of output produced by all intermediate
firms, including itself. Here, we assume that Xt(i) is defined by
Xt(i) = Z 1 0 Xt(i, j)(θ−1)/θdj θ/(θ−1) ,
where Xt(i, j) is firm i’s demand for the product j. Notice that the CES technology is exactly
the same as the final-good aggregator uses. Thus, it is equivalent to use the final good as
material inputs. The first-order condition for the cost-minimization problem of total material
costs implies that firm i’s inverse conditional demand function for intermediate product j is
given by Xt(i, j) = PX t (j) PX t −θ Xt(i), where PX
t (j) is the price of intermediate product j and P X t = h R1 0 Pt(j) 1−θdji1/(1−θ). The
aggregate demand for firm j’s product as materials inputs is then given by
Xt(j) = Z 1 0 Xt(i, j)di = PX t (j) PX t −θ Xt, where Xt = R1 0 Xt(i)di.
Similarly, Tt(i) is the CES aggregate of transportation services supplied by all
trans-portation firms, that is,
Tt(i) = Z 1 0 Tt(i, j)(θT−1)/θTdj θT/(θT−1) ,
where Tt(i, j) is firm i’s demand for type-j transportation services and θT > 1.
low of motion
Kt+1(i) = (1 − δt(i))Kt(i) + It(i), (5)
where It(i) is firm i’s investment at time t and δt(i) is time-varying depreciation rate. As
standard in the variable capital utilization literature (for example, Greenwood, Hercowitz,
and Huffman, 1988; Burnside and Eichenbaum, 1996), the capital depreciation rate δt(i)
depends on the rate of capital utilization through
δt(i) =
(ut(i))κ
κ , (6)
with κ > 1. As in Finn (2000), Leduc and Sill (2004), and Cavallo and Wu (2007), we
assume that capital utilization is tied with use of energy. Higher capital utilization must be
accompanied by more petroleum consumption per unit of capital, at an increasing rate. The
relationship is specified as EI t(i) Kt(i) = (ut(i)) ν ν , (7) where EI
t(i) represents firm i’s oil consumption for capital utilization and ν > 1. In order for
capital stock to be productive, each firm must purchase the necessary amount of petroleum
from the oil producer at the nominal price of PE t .
With the production technology (2) and (3), the optimality condition implies that
e Yt(i) = Qt(i) 1 − χY = Tt(i) χY and Qt(i) = Vt(i) 1 − χQ = Xt(i) χQ . (8)
Since each intermediate firm sells its distinct product to the final-good aggregator and other
firm i faces is given by Yt(i) + Xt(i) = Pt(i) Pt −θ Yt+ PX t (i) PX t −θ Xt = Pt(i) Pt −θ (Yt+ Xt) (9)
The last equality is due to the fact that there are no distinctions between outputs sold to
the final-good aggregator and those to intermediate-good firms as materials inputs, so that
Pt(i) = PtX(i) for all i ∈ [0, 1] and hence Pt= PtX. Each firm will make a decision such that
it can meet the total demand from the final-good aggregator and other intermediate-good
firms.
As in Ireland (2001), each intermediate-good firm is subject to two types of quadratic
adjustment costs for price and capital, which are respectively given by
ACY t(i) = φY 2 Pt(i) πsPt−1(i) − 1 2 Yt, (10) ACK t (i) = φK 2 Kt+1(i) Kt(i) − 1 2 Kt(i), (11)
where φY > 0 and φK > 0 are adjustment cost parameters and πs is the steady-state rate of
inflation.
Given (4)–(11), each intermediate firm chooses Pt(i), ut(i), Kt+1(i) and NtI(i) in order
to maximize its discounted market value defined as
E0 ∞ X t=0 βtΛt Pt Πt(i),
where Λt is the Lagrange multiplier associated with the household budget constraint (1),
Πt Pt = Pt(i) Pt e Yt(i) − wtNtI(i) − PE t Pt EI t(i) − Xt(i) − PT t Pt
Tt(i) − It(i) − ACtK(i) − AC Y t (i),
PE
t is the nominal price of oil, and P T
3.4
Transportation Firms
There is a continuum of transportation firms indexed by n ∈ [0, 1]. Each transportation firm
provides distinct transportation services Tt(n) in a monopolistically competitive market.
If there were no regulations, the transportation industry could be considered as a good
example of competitive industry. However, due to the regulations described in Section 2,
transportation firms are monopolistically competitive firms supplying distinctive services.
For example, we can think of regional barriers in the industry as one of the sources for
creating the distinctiveness.
Transportation services are produced by combining value added, which requires capital
and labor, and oil with a fixed-coefficient technology
Tt(n) = min [VtT(n), ωE T
t(n)] , (12)
where VT
t (n) is transportation firm n’s value added, which will be defined below, EtT(n) is
petroleum used for transportation, and ω > 0 is a transportation energy efficiency parameter.
Given (12), the optimality condition implies that
Tt(n) = VtT(n) = ωE T
t(n) (13)
The value-added production function of transportation firm n is assumed to be
VT
t (n) =aAt(n)(γ−1)/γ + (1 − a)NtT(n)
(γ−1)/γγ/(γ−1)
, (14)
where At(n) is transportation firm n’s beginning-of-period stock of transportation equipment,
NT
t (n) is labor inputs, a ∈ (0, 1), and γ is the elasticity of substitution between capital
and labor. Unlike the value-added production function of intermediate-good firms, that of
transportation firms takes the form of the CES production function. It is intended to take
value-added production of transportation with γ < 1, compared with that of
intermediate-good firms. Such complementarity can arise due to the fact that operating transportation
equipment requires certain amount of labor. I have also experimented with a typical
Cobb-Douglas value-added production, but the quantitative results on accounting for the changes
in the economy overtime are not significantly different.
Each transportation firm accumulates stock of transportation equipment following the
low of motion
At+1(n) = (1 − δA)At+ ItA(n), (15)
where δA is the time-invariant depreciation rate of At(n) and ItA(n) is transportation firm n’s
investment. Like intermediate-good producers, each transportation firm is subject to capital
adjustment costs, given by
ACA t(n) = φA 2 At+1(n) At(n) − 1 2 At(n), (16) with φA > 0.
Transportation firm n faces its own downward-sloping demand curve, which is given by
Tt(n) = PT t (n) PT t −θT Tt, (17) where PT t (n) is the price of Tt(n), PtT = h R1 0 P T t (n)1−θTdn i1/(1−θT) and Tt= R1 0 Tt(n)dn.
Subject to (14) – (17), each transportation firm chooses PT
t (n), N T
t (n), and At+1(n) to
maximize its discounted market value
E0 ∞ X t=0 βtΛt Pt ΠT t(n),
where ΠT t(n) Pt = P T t (n) Pt Tt(n) − wtNtT(n) − PE t Pt ET t(n) − I A t(n) − AC A t(n).
3.5
The Monetary Authority
Following Ireland (2001), it is assumed that the monetary authority will set short-term
interest rate rt following a Taylor-type policy rule
log(rt/rs) = ρylog(Yt/Ys) + ρπlog(πt/πs) + ρµlog(µt/µs) + εr,t,
where µt = MMt
t−1 and those variables with s-subscripts indicate the associated steady-state
values. Unlike the original policy rule proposed by Taylor (1993), this policy rule allows
monetary policy to respond to changes in money growth as well. Although it is well known
that the variant of the original Taylor rule estimated using pre-1979 data results in
indeter-minacy (e.g., Clarida, Gal´ı, and Gertler, 2000), Ireland (2001) notes that this specification
makes model’s equilibrium unique, even under the pre-1979 estimates.
3.6
Real Price of Oil
The log real price of oil follows an exogenous AR(1) process:
log(pE
t) = (1 − ρpE) log(pEs) + ρpElog(pEt−1) + εpE,t, (18)
where pE t = P
E
t /Pt, pEs represents the steady-state value the real price of oil, and |ρpE| < 1.
3.7
Symmetric Equilibrium
In a symmetric equilibrium, all agents in each sector make the same decision. The
NI t(n) = N I t and N T t(n) = N T
t for all i ∈ [0, 1], n ∈ [0, 1], and t. The market-clearing
condition of the labor market implies that Nt = NtI + N T
t . The market-clearing conditions
for money and bonds, Mt = Mt−1+ Tt and Bt= Bt−1 = 0, have to hold as well.
In the symmetric equilibrium, the household budget constraint (1) can be expressed as
Yt− pEt(E I t + E T t) = Ct+ It+ ItA+ AC W t + AC K t + AC Y t + AC A t. (19)
The equation (19) states that the sum of consumption, total investment, and total adjustment
costs equal to real value added, which is the value of output produced net of all costs
(including energy costs), except for primary-factor payments. Thus, an appropriate measure
of output in the model economy that corresponds to real GDP is real value added Yt −
pE t(E I t + E T t), rather than Yt.
The system of difference equations that characterize the symmetric equilibrium consists
of the first-order conditions for the households, the intermediate-good firms, and the
trans-portation firms, the laws of motion for capital, the monetary policy rule, and the oil-price
process.8
4
Results
This section presents the quantitative results and discusses how the model presented in
Section 3 responds to an oil-price shock. In order to obtain the solution to the system, I
log-linearize the equilibrium conditions of the model around its steady state and use the
generalized Schur decomposition described by Klein (2000) and Sims (2002).9 Before tuning
to showing the quantitative results, I describe the calibration procedure in the following
subsection.
8A complete description of the system is summarized in the Appendix, which is available from the author
upon request.
9
Table 1: Summary of Parameter Values in the Baseline Specification
Parameter Values
β Discount factor 0.9974∗
χQ Materials cost share 0.47324
χY Transportation share 0.0386
σ Money utility parameter 10
ϕ Frisch labor supply elasticity 1
α Capital share (intermediate-good firms) 0.36 a Capital share (transportation firms) 0.36 γ Transportation firms’ elasticity of substitution 0.1 δA Depreciation rate for transportation firms 0.025 φA Capital adjustment cost parameter (transportation firms) 10 φK Capital adjustment cost parameter (intermediate-good firms) 10
φW Wage adjustment cost parameter 58
φY Price adjustment cost parameter 72.01∗ θ Demand elasticity for consumption goods 6
θN Demand elasticity for labor 6
θT Demand elasticity for transportation services 3.8169 ρpE Persistence parameter of the oil price process 0.99 ρπ Monetary policy coefficient on inflation rate 0.8617∗ ρy Monetary policy coefficient on output 0.0499∗ ρµ Monetary policy coefficient on money growth 0.7351∗ pE
s Steady-state real price of oil 1
πs Steady-state inflation rate 1.0129∗
δs Steady-state depreciation rate for intermediate-good firms 0.025 s1 Steady-state value-added oil expenditure share for capital utilization 0.0123 s2 Steady-state value-added oil expenditure share for transportation 0.0242 ω Energy efficiency parameter for transportation 3.0334 ν Energy efficiency parameter for capital utilization 1.7373
κ Elasticity of depreciation 1.06
Note: Parameter values with asterisks are those based on the pre-1979 estimates of Ireland (2001).
4.1
Calibration
In order to analyze changes in the economy over time, I need to set the baseline specification
such that the model economy corresponds approximately to the state of the pre-1984 period.
In order to do so, I use pre-1984 values for some of share parameters, such as the value-added
oil expenditure shares. Furthermore, if available, I will utilize some of pre-1979 estimates
for parameter values from Ireland (2001) as an approximation. Parameter values assigned
Those parameter values drawn from the pre-1979 estimates of Ireland (2001) are a
discount factor β, the steady-state rate of inflation πs, the price adjustment cost parameter
φY and coefficients on the monetary policy rule, ρy, ρπ, and ρµ. β is set equal to 0.9974. With
the steady-state quarterly inflation rate πs = 1.0129, the value of β implies that annualized
interest rate is 6.3%.
Frisch labor supply elasticity is set to be unity. The value of σ is assumed to be 10, which
is broadly consistent with various estimates presented in Christiano, Eichenbaum, and Evans
(2005) with different specification.10 Capital share parameters α and a are set equal to 0.36.
The steady-state depreciation rate for intermediate-good firms δs and the time-invariant
depreciation rate for transportation firms δA are both equal to 0.025. Capital adjustment
cost parameter is assumed to be 10 for both intermediate-good firms and transportation
firms. The wage adjustment cost parameter is 58.11 Demand elasticity parameters for the
final good and for the index of labor services, θ and θN, respectively, equal to 6, which implies
in the steady-state, price and wage markup rates are 20%.
For parameter values related to transportation firms, I set the demand elasticity
param-eter θT to be 3.8169 for the baseline specification, such that the implied markup is 35.5%
and for the value after the deregulation in the transportation industry, I set it to 53.632, so
that the implied markup becomes 1.9%. The markup rate is consistent with the estimates
of Nebesky, McMullen, and Lee (1995).
Based on the industry-level data compiled by Dale Jorgenson and his colleagues,12 the
average share of materials cost (excluding energy cost) in the value of gross output from 1958
10Levin, Onatski, Williams, and Williams (2006) also have a similar estimate with Bayesian approach.
Although their estimate is about 11. This difference in the value of σ does not alter the results presented below.
11Wage rigidities introduced by the quadratic wage adjustment cost will have the same log-linearized
wage Phillips curve as Calvo-type wage stickiness. This value of wage adjustment cost parameter roughly corresponds to about 20% of households being able to reoptimize their wages in each period, given other parameter values used in the baseline specification.
12Data were downloaded from http://post.economics.harvard.edu/faculty/jorgenson/data/35klem.html.
Description of the data set is found in Jorgenson, Gollap, and Fraumeni (1987), Jorgenson and Stiroh (2000), and Jorgenson (1990).
to 1983 is 47.3%, and hence, the materials cost share χQ is set equal to 0.473.13 The value
of χY can be pinned down by the steady-state share of the transportation industry in the
total aggregate gross output from the optimality condition (8). The pre-1984 average share
of the transportation industry in Jorgenson’s data is 3.86%.14 Thus, I set χ
Y = 0.0386.
We set the steady-state value of the relative price of oil pE
s = 1 and the AR(1) coefficient
for the oil price process ρpE = 0.99. This is obtained from estimating (18) using the pre-1984
data on log-real price of oil, which is measured as log of PPI for crude petroleum deflated
by GDP deflator.
In terms of percentage deviations of the aggregate value added from the steady-state
value, the oil expenditure shares in aggregate value added play an important role. According
to the petroleum consumption data of the EIA and time series data on the price of
West-Texas Intermediate from FRED of the FRB St. Louis, the average share of oil consumption
in value added from 1949 to 1983 is 3.65%, and the average share after 1983 is 2.75%.
According to the EIA’s data, the average value of the ratio EI
/ET
is 0.51 in the pre-1984
period and 0.39 after 1984. Therefore, I set s1 = pEsE I s/ {Ys− pEs(E I s + E T s)} = 0.0123 and s2 = pEsE T s/ {Ys− pEs(E I s + E T
s)} = 0.0242 for the baseline specification.
Together with other parameter values, these calibrating targets on oil expenditure shares
are used to determine the remaining parameter values, which are related to energy efficiency
in the model economy, ω, ν, and κ. From the steady-state relation, the value of ω is given
by ω = (1 + s1+ s2) 1 + s1 s2 χYp E s (s1+ s2) {1 − χQ(1 − χY)} . (20)
The values of ν and κ are jointly determined by solving the following non-linear system of
13This number is close to the materials share used in Dotsey and King (2006). Although other authors such
as Rotemberg and Woodford (1996), and Huang and Liu (2004), use larger value for the share of materials, this is primarily because they take account of the energy cost as well.
equations: s1 s2 = (us) ν ν 1 β − (1 − δs) + p E s (us)ν ν −1 ΞY s Λs (1 − χQ(1 − χY)) αχVω χY , (21) 1 β − 1 = (us) κ 1 − 1 κ + pE s(us)ν 1 − 1 ν , (22) where us = (κδs)1/κ, Ξ Y s
Λs is the real marginal cost of producing an additional unit of V , and
χV =
(1−χQ)(1−χY)
1−χQ(1−χY) . Under the baseline specification, ω = 3.0334, ν = 1.7373, and κ = 1.06.
4.2
Baseline Results
First, we will look at the baseline specification, which corresponds to the pre-1984 state of
the economy. Figure 5 shows dynamic responses of variables to a 10% increase in the real
price of oil.
Aggregate value added, which corresponds to real GDP in data, shows a hump-shaped
response and peaks at 5 quarters after the shock. The near-unit-root nature of the oil shock
results in strong persistence in IRFs. Other variables, except for inflation, show hump-shaped
response as well, due to the price stickiness. Among macroeconomic variables, investment
is most severely affected, as pointed out by Aguiar-Conraria and Wen (2007). Two types of
oil consumption show different patterns in terms of magnitude. Oil consumption for capital
utilization is more price elastic compared with oil used for transportation. Aggregate hours
worked have a positive response in the first period due to the fact that the income effect
slightly dominates the substitution effect, and then decline. Under the monetary policy
rule assumed, the Fed increases interest rate in order to fight against inflation triggered by
increases in the oil price. Inflation dynamics look like a mirror image of the oil-price process
because there are no mechanisms that generate the hump-shaped response of inflation, such
as backward indexation or price adjustment costs for inflation.
There are two channels through which the oil-price shocks can affect the economy. The
0 4 8 12 16 20 24 28 −1.1 −1 −0.9 −0.8 −0.7
Aggregate Value Added
0 4 8 12 16 20 24 28 −0.8 −0.75 −0.7 −0.65 Consumption 0 4 8 12 16 20 24 28 −2 −1.5 −1 Aggregate Investment 0 4 8 12 16 20 24 28 −1.6 −1.4 −1.2 −1 Capital Utilization 0 4 8 12 16 20 24 28 −3 −2.5 −2 Oil (Utilization)
Percent deviations from the steady state
0 4 8 12 16 20 24 28 −0.8 −0.7 −0.6 −0.5 −0.4 Oil (Transportation) 0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 Real Wage 0 4 8 12 16 20 24 28 −0.3 −0.2 −0.1 0 Aggregate Hours 0 4 8 12 16 20 24 28 0.04 0.05 0.06 0.07 0.08 Interest Rate
Periods after the shock
0 4 8 12 16 20 24 28 0.06 0.08 0.1 0.12 0.14 0.16 Inflation
Periods after the shock
cost of utilizing capital stock, and hence a lower rate of capital utilization. This in turn
reduces the level of output produced. The second channel is a direct impact on the cost of
delivering goods produced.
Intermediate firms’ real marginal cost of producing an additional unit of output in terms
of log-deviations from the steady state is given by
c mct= 1 mcs (1 − χQ)(1 − χY)mc V smcc V t + χYp T spb T t , (23) where mcV s = ΞYs
Λs represents the steady-state real marginal cost of producing an additional
unit of their value added, pT
s is the real price of transportation services, and hat-variables
represent log-deviations from the steady state. Holding everything else constant, an increase
in the real price of oil affects bothmccVt via capital utilization and pbT
t through transportation
costs. Transportation firms’ markup plays a role in terms of amplifying the responses of
macroeconomic variables. The higher the markup, the more expensive the steady-state
price of transportation services. This is one of reasons why we expect the lower markup on
transportation services can contribute to the weaker response to the same size of the oil-price
shock.
Now, we turn to analyze factors that contribute to the observed weaker responses of
macroeconomic variables to the oil-price shock.
4.3
Changes in the Economy Over Time
In order to understand the declining effects of the oil-price shocks, we will consider three
different factors individually and some combinations of those. First, we will look at the
consequence of more competitive environment in the transportation industry triggered by
the deregulation after 1980. Second, we will see the effect of improved energy efficiency in
the economy. Lastly, we will look at the consequence of the less persistent oil-price shock.
Table 2: Summary of Changes in the Peak Response of Aggregate Value Added
% Change in Changes in Peak Responses the Timing
Case 1 Lower Markup −11.41% ±0
Case 2 Transportation Industry −19.73% ±0 Case 3 Improved Energy Efficiency −24.72% ±0 Case 4 Less Persistent Shock −26.52% −2
Case 5 Post-1984 −51.38% −2
each scenario and associated IRFs are shown in Figure 6 – Figure 8.
Case 1: Increased Competition among Transportation Firms
In order to see the outcome of the increased competition in the transportation industry,
we will only change the value of θT from 3.8169 to 53.632 such that the implied steady-state
markup of transportation firms changes from 35.5% to 1.9%, which is based on Nebesky,
McMullen, and Lee (1995). Other parameter values are left unchanged.
As shown in Figure 6 with the label of Lower Markup, all variables show weaker responses
to the same size of the oil-price shock. The peak response of aggregate value added is reduced
by 11.41%. However, there are no changes in the peak response.
Holding other things constant, the immediate effect of the lower markup is a reduction
in the steady-state price of transportation services pT
s. This induces a downward shift in the
marginal cost curve of intermediate-good firms. As a result, the economy moves to the new
steady state with the higher level of output and the lower level of the general price index. To
meet with the higher level of output, all factor inputs will be utilized more than before. It
should be noted that the steady-state real marginal cost of intermediate-good firms is equal
to (θ − 1)/θ, regardless of the value of θT.
Although the lower markup does not change the dynamics of pbT
t , the log-deviations of
the real marginal cost (23) tells us that a decline in pT
s induces in the deviations in the real
marginal cost, by putting less weight on deviations on the price of transportation firms. In
0 4 8 12 16 20 24 28 −1
−0.8 −0.6
Aggregate Value Added
0 4 8 12 16 20 24 28 −0.8 −0.7 −0.6 Consumption 0 4 8 12 16 20 24 28 −2 −1.5 −1 Aggregate Investment 0 4 8 12 16 20 24 28 −1.6 −1.4 −1.2 −1 Capital Utilization 0 4 8 12 16 20 24 28 −3 −2.5 −2 −1.5 Oil (Utilization)
Percent deviations from the steady state
0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 Oil (Transportation) 0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 −0.2 Real Wage 0 4 8 12 16 20 24 28 −0.3 −0.2 −0.1 0 0.1 Aggregate Hours 0 4 8 12 16 20 24 28 0.04 0.06 0.08 Interest Rate
Periods after the shock
0 4 8 12 16 20 24 28 0.05
0.1 0.15
Inflation
Periods after the shock
Baseline Lower Markup Deregulation
Figure 6: Comparison of IRFs to a 10% Increase in the Oil Price: Case 1 and Case 2
We can observe another effect of lowering the markup on the inflation dynamics. The
equi-librium is given by 0 = (1 − θ)Yt+ θ {χQ(1 − χY) + χYp T t} Yt− φY πt πs − 1 πt πs (1 − χQ(1 − χY)) Yt + θΞ Y t Λt (1 − χQ)(1 − χY)Yt+ βEt Λt+1 Λt φY πt+1 πs − 1 πt+1 πs (1 − χQ(1 − χY)) Yt+1 (24)
Log-linearizing the above equation gives us the New-Keynesian-Phillips-curve type relation:
b πt= θ φY (1 − χQ)(1 − χY)mc V s 1 − χQ(1 − χY) c mcVt + χYp T s 1 − χQ(1 − χY) b pT t + βEt[πbt+1]. (25)
Thus, immediately from the above equation, a reduction in pT
s induces smaller deviations in
inflation.
Case 2: Transportation Industry (Deregulation)
As described in Section 2, another consequence of the deregulation is that it has
con-tributed to improve efficiency in the transportation industry. It is important to consider this
contribution of the transportation industry as a whole because it accounts for the majority
of US petroleum consumption. Here we will consider the contribution of the transportation
industry, by taking account of the lower markup and the improved energy efficiency in the
industry.
Unfortunately, we do not have reliable data on how energy efficiency has been improved
in the transportation industry. One thing we know is that the average oil expenditure share
has dropped from 3.65% in the pre-1984 period to 2.75% in the post-1984 period. Another
thing we can observe is changes in petroleum consumption in industrial and transportation
sectors. Particularly, the ratio s1/s2 has decreased from 0.51 to 0.39 in the post-1984 period.
consistent with the post-1984 characteristic of the US petroleum consumption, by using the
above information. As a result, ω increases from 3.0334 to 3.6752.
In order to see the consequence of the deregulation, I will change both θT and ω from
the values used in the baseline specification. Since improved energy efficiency in the
indus-try could be because of the deregulation and/or advanced transportation technology, such
as higher miles per gallon, this specification does not represent the exact outcome of the
deregulation. I sill believe that it is a good proxy for the consequence of the deregulation,
however. If there still existed restrictions on entry/exit in the industry, then there would
be no entrants that are more efficient and there would be less incentive for incumbents to
improve efficiency by adopting more fuel efficient technology.
IRFs presented in Figure 6 labeled Deregulation report the contribution originated from
the transportation sector of the economy. All variables show weaker responses to the same
size of the shock than the baseline specification. In terms of changes in the aggregate
value-added peak responses, the consequence of the deregulation accounts for a 19.73% reduction,
compared to the baseline specification.
Improved energy efficiency in the transportation sector, that is, an increase in ω has
two effects. First, it reduces the steady-state price of transportation services, which has
the same effects as the lower markup. Under this specification, the steady-state price of
transportation services decreases by 31%, which is consistent with the empirical finding of
Ying and Keeler (1991). Second, it makes deviations of the transportation price smaller.
b pT t = 1 pT s mcT smcc T t + pE s ωpb E t (26)
It is clear that an increase in ω makes pbT
t smaller than before. In turn, it makes deviations
in the marginal cost smaller.
A 20% increase in ω used in this specification is much smaller than the 42%
because of the fact that the US economy is more dependent on transportation, which is
apparent in a large drop in the ratio s1/s2. If the ratio stayed the same, which is equivalent
to less reliance on transportation, we would expect a bigger contribution from the
trans-portation industry.
Case 3: Improved Energy Efficiency in the Economy
Another factor that could contribute to the weaker responses of macroeconomic variables
is improvements in efficiency of petroleum use. It is quite intuitive that improved energy
efficiency might be a candidate for the declining effects of the oil-price shock. As energy
efficiency improves, the economy needs to rely less on petroleum and becomes less vulnerable
to an increase in the oil price. This is apparent in the smaller oil expenditure share.
More efficient use of oil could originate from capital utilization and/or transportation.
Thus, here I will change both ω and ν, which govern energy efficiency in both transportation
and industrial sectors, respectively. Given the value of ω used in Case 2 and other parameter
values fixed, ν is changed such that the resulting steady-state oil expenditure share equals
to 2.75%. As a result, ν changes from 1.7373 to 1.8864.
Given the same size of shock, IRFs in Figure 7 labeled Energy Efficiency also show
weaker responses of all variables. As a result, the peak response of aggregate value added
becomes smaller by 24.72%, compared with the baseline specification.
An increase in ν has an effect similar to an increase in ω. It results in the smaller value
of steady-state marginal cost of producing intermediate-good fimrs’ value added mcV s and
also mccVt becomes smaller given the same size of shock in the oil price.
Case 4: Less Persistence of the Shock
It is not hard to imagine that the shape of IRFs would be affected by a small change in
the underlying shock process, and thus one might suspect that it could be the whole story
0 4 8 12 16 20 24 28 −1 −0.8 −0.6 −0.4 −0.2
Aggregate Value Added
0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 −0.2 Consumption 0 4 8 12 16 20 24 28 −2 −1.5 −1 −0.5 Aggregate Investment 0 4 8 12 16 20 24 28 −1.5 −1 −0.5 0 Capital Utilization 0 4 8 12 16 20 24 28 −3 −2 −1 Oil (Utilization)
Percent deviations from the steady state
0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 −0.2 Oil (Transportation) 0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 −0.2 Real Wage 0 4 8 12 16 20 24 28 −0.3 −0.2 −0.1 0 0.1 Aggregate Hours 0 4 8 12 16 20 24 28 0.02 0.04 0.06 0.08 Interest Rate
Periods after the shock
0 4 8 12 16 20 24 28 0.05
0.1 0.15
Inflation
Periods after the shock
Baseline Energy Efficiency Less Persistence
0 4 8 12 16 20 24 28 −1 −0.8 −0.6 −0.4 −0.2
Aggregate Value Added
0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 −0.2 Consumption 0 4 8 12 16 20 24 28 −2 −1.5 −1 −0.5 Aggregate Investment 0 4 8 12 16 20 24 28 −1.5 −1 −0.5 0 Capital Utilization 0 4 8 12 16 20 24 28 −3 −2 −1 Oil (Utilization)
Percent deviations from the steady state
0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 −0.2 Oil (Transportation) 0 4 8 12 16 20 24 28 −0.8 −0.6 −0.4 −0.2 Real Wage 0 4 8 12 16 20 24 28 −0.3 −0.2 −0.1 0 0.1 Aggregate Hours 0 4 8 12 16 20 24 28 0.02 0.04 0.06 0.08 Interest Rate
Periods after the shock
0 4 8 12 16 20 24 28 0.05
0.1 0.15
Inflation
Periods after the shock
Baseline Post−1984
persistence in the oil price is easily observed. In the post-1984 period, the estimated AR(1)
coefficient on the log-real price, measured as log of PPI for crude petroleum deflated by
GDP deflator, suggests that it is dropped from 0.99 to 0.92. Thus, I change the persistence
parameter ρpE from 0.99 to 0.92.
IRFs shown in Figure 7 labeled Less Persistence are the resulting responses. In terms
of the peak response, as expected, a small change in the persistence parameter contributes
to a large reduction in the peak response, 26.52%. Although there are no changes in the
timing of the peak under other factors considered, less persistence in the oil-price process
shifts the timing of the peak by 2 quarters.
Case 5: Post-1984
Finally, I combine all factors considered in this paper together, in order to see how much
the model could explain changes in the economy over time in response to the oil-price shock.
Figure 8 compares resulting IRFs with those from the baseline specification. As can be seen
clearly, all factors together result in much smaller responses to the same size of the oil-price
shock. In terms of the changes in the peak response of aggregate value added, it suggests
that the peak response is reduced by 51.38% as a whole and the timing of the peak is shifted
by 2 quarters due to the less persistent oil-price process.
If one believes the VAR results presented in the Introduction are a good representation
of changes in the economy over time, then these factors considered in this paper could
account for a majority of the declining effects of the oil-price shock. It should be noted that,
however, these results do not suggest that the factors considered here are the only source of
changes in the economy over time, or that all changes have made the effects smaller. For
example, in the presence of the trade-off between output and inflation stabilization, a strong
anti-inflation stance of the monetary authority, which is widely documented, then it would
Fed could stabilize the output response as demonstrated in Blanchard and Gal´ı (2007).
Alternatively, one might argue that the economy is shifting toward more flexible prices.15
Regarding reduced price stickiness, it would contribute to shifting the timing of the peak.
The more flexible the price is, the earlier peak we would expect. Since price stickiness
has dampening effects on dynamic response of variables, more flexible price would result in
stronger responses of aggregate value added, however. Mechanically, less habit persistence, if
any, could work in the same way as flexible prices. As argued in Blanchard and Gal´ı (2007),
it could be also true that wage rigidities have become smaller over time.
To sum up, by comparing Case 1 and Case 2, we can see that two factors affecting
the contribution of the transportation industry, namely more competition and more efficient
use of oil in the industry, are equally important. The effect of the deregulation in the
transportation industry results in a 20% reduction in the peak response of aggregate value
added. It should be noted that in spite of its small share in the economy, the contribution
of the transportation industry is non-negligible. The overall effect of more efficient use of
oil in the US economy weakens the peak responses of value added by 25%. The results
suggest that the industrial sector (intermediate-good firms) has a slightly larger impact on
the weaker response than the transportation sector does, since improved energy efficiency in
the transportation industry roughly accounts for about a 10% reduction in the peak response.
A less persistent oil-price shock accounts for a 27% reduction in the peak response. Overall,
all factors play an important role in accounting for the declining effects of the oil-price shock
and are equally important quantitatively.
5
Conclusion
We have developed a model that is more realistic in terms of US petroleum consumption, in
order to explore possible explanations for the weaker responses of macroeconomic variables
15It has been widely believed that the average frequency of price change is about one year (for example,
to oil-price shocks. Under the baseline specification, which corresponds to the pre-1984 state
of the economy, the model well captures important aspects of the economy’s response to the
oil-price shock. Price stickiness accounts for hump-shaped responses of output measures,
which are consistent with what we see in data.
We have explored three developments that may have affected the response of output
to an oil-price shock, namely the effect of deregulation in the transportation industry,
im-proved energy efficiency, and less persistent oil-price shocks. All factors are equally important
quantitatively. It is worth noting that in spite of its small share, the contribution of the
trans-portation industry triggered by the deregulation is nontrivial. If the US economy becomes
less dependent on transportation, we would expect further contribution of the
transporta-tion sector. Combining all factors together, the model predicts a 51% reductransporta-tion in the peak
response of aggregate value added and the peak comes 2-quarter earlier.
We can draw one implication from this study. Deregulation in the transportation
indus-try has led to more competitive environment and less reliance on petroleum. Also,
techno-logical advancement enables us more efficient use of petroleum overall. Both factors should
make the economy less vulnerable to oil-price shocks in the future. From past experience,
the general public as well as policy makers might still hold a view that large increases in the
price of oil trigger a deep recession. However, we expect that large recessionary consequences